[llvm] [AMDGPU] Change the representation of double literals in operands (PR #68740)

[Stanislav Mekhanoshin via llvm-commits]

================
@@ -454,6 +454,8 @@ void AMDGPUInstPrinter::printImmediate64(uint64_t Imm,
   else if (Imm == 0x3fc45f306dc9c882 &&
            STI.hasFeature(AMDGPU::FeatureInv2PiInlineImm))
     O << "0.15915494309189532";
+  else if (IsFP && AMDGPU::isValid32BitLiteral(Imm, true))
----------------
rampitec wrote:

Yes, it can, given that lo32 are cleared by the parser. Changed.

https://github.com/llvm/llvm-project/pull/68740